Would you publish this briefing as written?
Your output: A corrected brief and a claim-by-claim evidence record.
01 · The assignment
The setting is a fictional editorial team preparing a short briefing for small-business owners. The editor wants to explain a pilot study without overstating its findings.
The task is to check a generated draft against the evidence packet below. All names, source labels, and numbers in this exercise are invented.
The constraint is a short deadline, which makes a fluent draft tempting to approve. However, publication depends on supported claims, not polished wording.
02 · Inspect the evidence
Source A, a fictional pilot note, describes 12 volunteers completing a specific drafting task. Median drafting time was 20 minutes without assistance and 15 minutes with assistance.
Source B, a fictional review note, says the pilot did not measure total working time or long-term performance. It also says reviewers corrected unsupported details in several drafts.
Source C, a fictional bibliography record, matches the pilot title, author, and year. That match identifies the document; it does not independently verify its findings.
| Source | What it can establish | What it cannot establish |
|---|---|---|
| A · Pilot note | A difference in median drafting time for this task | A causal effect across all teams |
| B · Review note | Review work remained necessary | The amount of total time saved |
| C · Bibliography record | The identity of the cited document | Whether a sentence accurately represents the findings |
03 · Audit the draft
The generated draft says: “AI makes teams 25% more productive and eliminates review work. The pilot proves every business should adopt it.”
The first claim changes a drafting-time result into a productivity claim. A 25% reduction in one measured duration does not establish 25% more output.
The second claim contradicts the review note, while the third leaps beyond the sample. Therefore, the editor separates the sentence into independently checkable statements.
| Claim | Verdict | Editorial action |
|---|---|---|
| Teams are 25% more productive | Unsupported interpretation | Replace with the measured drafting-time comparison |
| Review work disappears | Contradicted by Source B | State that review remained necessary |
| Every business should adopt AI | Unsupported generalization | Remove the universal recommendation |
04 · Repair the reasoning
The calculation is (20 − 15) ÷ 20 × 100 = 25%. This describes the change in median drafting time in the fictional pilot.
The revised brief reads: “In a fictional pilot with 12 volunteers, median drafting time fell from 20 to 15 minutes. Review remained necessary, and total working time was not measured.”
The qualification belongs next to the finding, where readers can see it. Moving all limitations into a distant footnote would leave the main claim misleading.
05 · Make the editorial decision
The decision is to reject the original wording and accept the narrower revision for this exercise. The source packet supports the revised description, within its stated limits.
A real publication would require the actual source, relevant passages, and checks for corrections or conflicting evidence. Crossref metadata can help identify scholarly works, but a matching record is not proof of claim support.
The evidence record should preserve the original claim, relevant passage, verdict, correction, and reviewer. That record makes a later update easier when new information arrives.
06 · Transfer the method
Your challenge is to rewrite the brief for a reader who only sees two sentences. Preserve the measured outcome, the sample limit, and the need for review.
The counterfactual adds a five-minute increase in review time per task. Explain why you cannot subtract that figure directly from medians without suitable paired data.
Success looks like a revision whose certainty matches the evidence. You can distinguish source identity, claim support, and scope of inference.
Make your decision first.
Write your recommendation and the evidence that supports it. Then compare your reasoning with the worked answer below.
Reveal the worked decision
Reject the original. The packet supports a 25% reduction in median drafting time for one fictional task, with review still required. It does not support higher overall productivity, universal adoption, or removal of review. If review time rises, assess total task time with appropriate measurements instead of combining incompatible summaries.
What a strong response includes
- Evidence fidelity: Every factual statement traces to the packet.
- Scope: The wording names the measured task and sample limits.
- Numeracy: The percentage describes time reduction, not a different metric.
- Editorial judgment: Unsupported recommendations are removed or explicitly labeled.
Take the method into your work.
Download the editable Markdown worksheet. Open it in a text editor and replace the blank sections with fictional or approved material.
Download the worksheet (.md) ↓